Generally, image signals obtained from an imaging system comprising imaging devices and their accompanying analog circuits, A/D converters, etc. contain noise components. The noise components can be categorized roughly into fixed pattern noise and random noise. The fixed pattern noise is noise caused mainly by the imaging devices such as defective pixels. On the other hand, the random noise is generated in the imaging devices and the analog circuits, and has characteristics close to white noise.
Regarding reduction processing of random noise in a still image, for example, as shown in JP2001-175843A, a technique of separating an image signal into a luminance signal and a color difference signal, obtaining edge intensity from the luminance signal and the color difference signal, and performing smoothing processing to the color difference signal in a region other than the edge region is disclosed. Thereby, it becomes possible to perform color noise reduction processing to only the smoothed portion while reserving the edge portion.
Further, as shown in JP2005-175718A, an example of separating an image signal into a luminance signal and a color difference signal, estimating a luminance noise amount and a color noise amount for each predetermined region, and performing noise reduction processing for each region is disclosed. Thereby, it becomes possible to perform optimized noise reduction processing for each region.
On the other hand, as an example of noise reduction processing of random noise in motion pictures, there is recursive type noise reduction processing using correlation characteristics in a direction of time axis. In the recursive type noise reduction processing, based on such a fact that the image signal itself has high correlation with the image signal of the past while the random noise has low correlation with the image signals of the past, only random noise is extracted by performing differential processing between the present and past image signals, and noise reduction processing for the present image signal is performed using the extracted random noise.
As such recursive type noise reduction processing, a method for detecting a motion component from an image signal and controlling a limit value and a feedback coefficient for the signal which has been subjected to differential processing based upon the detected motion component is disclosed in JP10-13734A. Thereby, it becomes possible to perform noise reduction processing where adverse effect such as afterimage due to the motion component has been suppressed. Moreover, JP2000-209507A discloses a method for controlling a feedback coefficient from a signal value which has been subjected to differential processing. Thereby, it becomes possible to perform noise reduction processing where adverse effect has been suppressed with relatively reduced arithmetic amount.